Viral AI image trends drive up water consumption

Behind ChatGPT’s digital charm lies an increasingly concerning environmental toll, largely driven by its water consumption.

According to recent reports, OpenAI’s GPT-4 model consumes around 500 millilitres of clean, drinkable water for every 100-word response. The surge in demand, fuelled by viral trends like Studio Ghibli-style portraits and Barbie-themed avatars, has significantly amplified this impact.

Each AI interaction, especially those involving image generation, generates heat, necessitating cooling systems that rely heavily on water.

With an estimated 57 million users daily, ChatGPT’s operations result in a staggering daily water usage of over 14,800 crore litres. OpenAI’s CEO, Sam Altman, recently acknowledged server strain, urging users to reduce non-essential use.

The environmental costs extend beyond water. Many data centres supporting AI platforms are located in water-stressed regions and rely on fossil fuels, raising serious concerns about sustainability.

Experts warn that while AI promises convenience, its rapid expansion risks putting additional pressure on fragile ecosystems unless mindful practices are adopted.

Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!

AI demand drives up water consumption

The growing demand for data centres, driven by the AI boom, is leading to a significant increase in water consumption, particularly for cooling the computing equipment. In Virginia, home to the world’s largest concentration of data centres, water usage surged by nearly two-thirds between 2019 and 2023, rising from 1.13 billion to 1.85 billion gallons.

The trend, mirrored globally, raises concerns about sustainability. Microsoft, a key player in the data centre industry, reported that 42% of the water it used in 2023 came from regions experiencing water stress. Google, which operates some of the largest data centres, revealed that 15% of its freshwater withdrawals occurred in areas with high water scarcity.

Although many data centres use closed-loop systems to recycle water, a significant portion is still lost due to the need for humidity control, especially in dry regions. Humidified air is essential to prevent static electricity, which can damage sensitive computer equipment.

The increasing water consumption by data centres underscores the environmental challenges posed by the rapid expansion of AI and digital infrastructure, prompting concerns about the sustainability of such practices.